Expand KOC LOOP MCP operations

This commit is contained in:
巫凤萍
2026-08-07 14:12:52 +08:00
parent 1a9a113229
commit 5670a7939b
15 changed files with 2426 additions and 24 deletions

75
lib/mcp-export-token.ts Normal file
View File

@@ -0,0 +1,75 @@
import { ensureSchema, getRawDb } from "./mvp-db";
export type McpExportKind = "recovery" | "resources";
type ExportTokenRow = {
kind: McpExportKind;
payload: string;
expires_at: string;
};
function sqliteTimestamp(date: Date) {
return date.toISOString().replace("T", " ").slice(0, 19);
}
async function digest(value: string) {
const bytes = await crypto.subtle.digest(
"SHA-256",
new TextEncoder().encode(value),
);
return [...new Uint8Array(bytes)]
.map((byte) => byte.toString(16).padStart(2, "0"))
.join("");
}
export async function issueMcpExportToken(
kind: McpExportKind,
payload: Record<string, unknown>,
ttlMinutes = 15,
) {
await ensureSchema();
const db = getRawDb();
const token = `${crypto.randomUUID().replaceAll("-", "")}${crypto
.randomUUID()
.replaceAll("-", "")}`;
const expiresAt = new Date(
Date.now() + Math.max(1, Math.min(60, ttlMinutes)) * 60_000,
);
await db.prepare("DELETE FROM mcp_export_tokens WHERE expires_at <= CURRENT_TIMESTAMP").run();
await db
.prepare(
`INSERT INTO mcp_export_tokens (token_hash, kind, payload, expires_at)
VALUES (?, ?, ?, ?)`,
)
.bind(await digest(token), kind, JSON.stringify(payload), sqliteTimestamp(expiresAt))
.run();
return { token, expiresAt: expiresAt.toISOString() };
}
export async function consumeMcpExportToken(
token: string,
expectedKind: McpExportKind,
) {
if (!token || token.length < 32) return null;
await ensureSchema();
const db = getRawDb();
const tokenHash = await digest(token);
const row = await db
.prepare(
`SELECT kind, payload, expires_at
FROM mcp_export_tokens
WHERE token_hash = ? AND expires_at > CURRENT_TIMESTAMP`,
)
.bind(tokenHash)
.first<ExportTokenRow>();
if (!row || row.kind !== expectedKind) return null;
await db
.prepare("DELETE FROM mcp_export_tokens WHERE token_hash = ?")
.bind(tokenHash)
.run();
try {
return JSON.parse(row.payload) as Record<string, unknown>;
} catch {
return null;
}
}

464
lib/mcp-operations.ts Normal file
View File

@@ -0,0 +1,464 @@
import { enrichDistributionAccount } from "./account-enrichment-service";
import {
collectDistributionMetrics,
createCollectionRunTasks,
retryFailedCollections,
runDueScheduledCollections,
shanghaiDateFromTimestamp,
} from "./collection-service";
import { issueMcpExportToken } from "./mcp-export-token";
import {
resolveCollectionMcpConfig,
type CollectionMcpBindings,
} from "./mcp-collection-client";
import { ensureSchema, getRawDb } from "./mvp-db";
import { extractXhsPublishUrl } from "./publish-url";
import { buildClaimUrl } from "./task-service";
export type McpOperationBindings = CollectionMcpBindings & {
KOC_PORTAL_URL?: string;
};
type Pagination = { limit?: number; offset?: number };
type ResourceFilters = Pagination & {
query?: string;
ipLocation?: string;
cooperationSource?: string;
platform?: string;
};
function page(input: Pagination) {
return {
limit: Math.max(1, Math.min(200, Math.floor(input.limit ?? 50))),
offset: Math.max(0, Math.floor(input.offset ?? 0)),
};
}
function like(value: string) {
return `%${value.replace(/[\\%_]/g, "\\$&")}%`;
}
function parseJsonArray(value: unknown) {
try {
const parsed = JSON.parse(String(value ?? "[]"));
return Array.isArray(parsed) ? parsed : [];
} catch {
return [];
}
}
async function taskExists(taskId: string) {
const task = await getRawDb()
.prepare("SELECT id FROM tasks WHERE id = ?")
.bind(taskId)
.first<{ id: string }>();
if (!task) throw new Error("任务不存在");
}
export async function taskList(
input: Pagination & { query?: string; status?: string },
portalUrl: string,
) {
await ensureSchema();
const { limit, offset } = page(input);
const conditions: string[] = [];
const bindings: unknown[] = [];
if (input.query?.trim()) {
conditions.push("(t.name LIKE ? ESCAPE '\\' OR t.brand LIKE ? ESCAPE '\\')");
const pattern = like(input.query.trim());
bindings.push(pattern, pattern);
}
if (input.status?.trim() && input.status !== "all") {
conditions.push("t.status = ?");
bindings.push(input.status.trim());
}
const where = conditions.length ? `WHERE ${conditions.join(" AND ")}` : "";
const db = getRawDb();
const [rows, count] = await Promise.all([
db
.prepare(
`SELECT t.*,
(SELECT COUNT(*) FROM contents c WHERE c.task_id = t.id) AS note_count,
(SELECT COUNT(*) FROM distributions d WHERE d.task_id = t.id) AS claimed_count,
(SELECT COUNT(*) FROM distributions d WHERE d.task_id = t.id AND COALESCE(d.publish_url, '') != '') AS published_count,
(SELECT COUNT(*) FROM distributions d WHERE d.task_id = t.id AND d.exposure IS NOT NULL AND d.views IS NOT NULL) AS day7_count
FROM tasks t ${where}
ORDER BY t.created_at DESC LIMIT ? OFFSET ?`,
)
.bind(...bindings, limit, offset)
.all<Record<string, unknown>>(),
db
.prepare(`SELECT COUNT(*) AS total FROM tasks t ${where}`)
.bind(...bindings)
.first<{ total: number }>(),
]);
return {
total: count?.total ?? 0,
limit,
offset,
tasks: rows.results.map((row) => ({
...row,
collection_days: parseJsonArray(row.collection_days),
claim_url:
portalUrl && row.share_token
? buildClaimUrl(portalUrl, String(row.share_token))
: null,
})),
};
}
export async function taskGet(taskId: string, portalUrl: string) {
await ensureSchema();
const db = getRawDb();
const task = await db
.prepare("SELECT * FROM tasks WHERE id = ?")
.bind(taskId)
.first<Record<string, unknown>>();
if (!task) throw new Error("任务不存在");
const [notes, claims, runs] = await Promise.all([
db
.prepare(
`SELECT c.id AS content_id, c.source_row, c.title, c.body, c.image_assets, c.status AS content_status,
d.id AS distribution_id, d.status AS distribution_status, d.publish_url, d.publish_time,
d.exposure, d.views, d.latest_likes, d.latest_comments, d.latest_collects,
d.collection_status, d.collection_status_description, d.collection_updated_at,
a.id AS account_id, a.nickname AS account_nickname, a.profile_url, a.followers,
p.name AS cooperation_source, cl.claimant_name
FROM contents c
LEFT JOIN distributions d ON d.content_id = c.id
LEFT JOIN accounts a ON a.id = d.account_id
LEFT JOIN partners p ON p.id = d.partner_id
LEFT JOIN claims cl ON cl.id = d.claim_id
WHERE c.task_id = ?
ORDER BY COALESCE(c.source_row, 999999), c.created_at, d.claimed_at`,
)
.bind(taskId)
.all<Record<string, unknown>>(),
db
.prepare(
`SELECT cl.*, p.name AS partner_name,
(SELECT COUNT(*) FROM distributions d WHERE d.claim_id = cl.id AND COALESCE(d.publish_url, '') != '') AS published_count
FROM claims cl JOIN partners p ON p.id = cl.partner_id
WHERE cl.task_id = ? ORDER BY cl.created_at DESC`,
)
.bind(taskId)
.all<Record<string, unknown>>(),
db
.prepare(
`SELECT * FROM collection_runs WHERE task_id = ?
ORDER BY scheduled_date DESC, created_at DESC LIMIT 500`,
)
.bind(taskId)
.all<Record<string, unknown>>(),
]);
return {
task: {
...task,
collection_days: parseJsonArray(task.collection_days),
claim_url:
portalUrl && task.share_token
? buildClaimUrl(portalUrl, String(task.share_token))
: null,
},
notes: notes.results.map((row) => ({
...row,
image_assets: parseJsonArray(row.image_assets),
total_interactions:
row.latest_likes == null
? null
: Number(row.latest_likes) +
Number(row.latest_comments ?? 0) +
Number(row.latest_collects ?? 0),
})),
claims: claims.results,
collection_runs: runs.results,
};
}
export async function recoveryList(
input: Pagination & {
taskId?: string;
stage?: "all" | "published" | "unfilled" | "waiting_day7" | "day7_due";
},
) {
await ensureSchema();
const { limit, offset } = page(input);
const db = getRawDb();
const conditions: string[] = [];
const bindings: unknown[] = [];
if (input.taskId?.trim()) {
conditions.push("d.task_id = ?");
bindings.push(input.taskId.trim());
}
const stage = input.stage ?? "all";
if (stage === "published") conditions.push("COALESCE(d.publish_url, '') != ''");
if (stage === "unfilled") conditions.push("COALESCE(d.publish_url, '') = ''");
if (stage === "waiting_day7") {
conditions.push("COALESCE(d.publish_url, '') != '' AND (d.exposure IS NULL OR d.views IS NULL)");
}
if (stage === "day7_due") {
conditions.push(
"COALESCE(d.publish_url, '') != '' AND (d.exposure IS NULL OR d.views IS NULL) AND datetime(d.publish_time, '+7 days') <= CURRENT_TIMESTAMP",
);
}
const where = conditions.length ? `WHERE ${conditions.join(" AND ")}` : "";
const base = `FROM distributions d
JOIN tasks t ON t.id = d.task_id
JOIN contents c ON c.id = d.content_id
LEFT JOIN accounts a ON a.id = d.account_id
LEFT JOIN partners p ON p.id = d.partner_id
LEFT JOIN claims cl ON cl.id = d.claim_id ${where}`;
const [rows, count] = await Promise.all([
db
.prepare(
`SELECT d.*, t.name AS task_name, c.source_row, c.title,
a.nickname AS account_nickname, a.profile_url, p.name AS cooperation_source,
cl.claimant_name ${base}
ORDER BY COALESCE(d.publish_time, d.claimed_at) DESC LIMIT ? OFFSET ?`,
)
.bind(...bindings, limit, offset)
.all<Record<string, unknown>>(),
db
.prepare(`SELECT COUNT(*) AS total ${base}`)
.bind(...bindings)
.first<{ total: number }>(),
]);
return {
total: count?.total ?? 0,
limit,
offset,
items: rows.results.map((row) => ({
...row,
total_interactions:
row.latest_likes == null
? null
: Number(row.latest_likes) + Number(row.latest_comments ?? 0) + Number(row.latest_collects ?? 0),
})),
};
}
export async function recoveryExport(taskId: string, origin: string) {
await ensureSchema();
await taskExists(taskId);
const issued = await issueMcpExportToken("recovery", { taskId });
return {
task_id: taskId,
download_url: `${origin}/api/recovery-export?token=${encodeURIComponent(issued.token)}`,
expires_at: issued.expiresAt,
};
}
export async function setCollectionPlan(
taskId: string,
startDate: string,
days: number[],
bindings: McpOperationBindings,
) {
await ensureSchema();
await taskExists(taskId);
const normalizedDays = [...new Set(days)]
.filter((day) => Number.isInteger(day) && day >= 1 && day <= 7)
.sort((a, b) => a - b);
if (!/^\d{4}-\d{2}-\d{2}$/.test(startDate) || Number.isNaN(Date.parse(`${startDate}T00:00:00+08:00`))) {
throw new Error("开始采集日期无效");
}
if (!normalizedDays.length) throw new Error("请至少选择一个采集日");
const db = getRawDb();
await db.batch([
db.prepare(
`UPDATE tasks SET collection_start_date = ?, collection_days = ?,
collection_schedule_updated_at = CURRENT_TIMESTAMP WHERE id = ?`,
).bind(startDate, JSON.stringify(normalizedDays), taskId),
db.prepare(
`UPDATE distributions SET
collection_status = CASE WHEN latest_likes IS NULL THEN 'scheduled' ELSE collection_status END,
collection_status_description = CASE WHEN latest_likes IS NULL THEN ? ELSE collection_status_description END,
updated_at = CURRENT_TIMESTAMP
WHERE task_id = ? AND COALESCE(publish_url, '') != ''`,
).bind(`已安排${normalizedDays.length}个采集日每日10:00执行`, taskId),
]);
const created = await createCollectionRunTasks(db, taskId, startDate, normalizedDays);
const catchup = await runDueScheduledCollections(
db,
Date.now(),
resolveCollectionMcpConfig(bindings),
"catchup",
taskId,
);
return { task_id: taskId, start_date: startDate, days: normalizedDays, created, catchup };
}
export async function runDueCollections(taskId: string | undefined, bindings: McpOperationBindings) {
await ensureSchema();
if (taskId) await taskExists(taskId);
return runDueScheduledCollections(
getRawDb(),
Date.now(),
resolveCollectionMcpConfig(bindings),
"catchup",
taskId,
);
}
export async function collectNow(
distributionId: string,
scheduleDay: number | undefined,
bindings: McpOperationBindings,
) {
await ensureSchema();
const day = scheduleDay ?? null;
const metrics = await collectDistributionMetrics(
getRawDb(),
distributionId,
shanghaiDateFromTimestamp(Date.now()),
day,
"manual",
resolveCollectionMcpConfig(bindings),
);
return { distribution_id: distributionId, schedule_day: day, ...metrics };
}
export async function retryFailed(taskId: string | undefined, bindings: McpOperationBindings) {
await ensureSchema();
const db = getRawDb();
const ids = taskId
? [taskId]
: (
await db
.prepare(
`SELECT DISTINCT task_id FROM distributions
WHERE collection_status = 'failed' ORDER BY updated_at DESC LIMIT 20`,
)
.all<{ task_id: string }>()
).results.map((row) => row.task_id);
if (taskId) await taskExists(taskId);
const results = [];
for (const id of ids) {
results.push({ task_id: id, ...(await retryFailedCollections(db, id, resolveCollectionMcpConfig(bindings))) });
}
return {
task_count: results.length,
attempted: results.reduce((sum, item) => sum + item.attempted, 0),
succeeded: results.reduce((sum, item) => sum + item.succeeded, 0),
failed: results.reduce((sum, item) => sum + item.failed, 0),
tasks: results,
};
}
function resourceWhere(input: ResourceFilters) {
const conditions: string[] = [];
const bindings: unknown[] = [];
if (input.query?.trim()) {
conditions.push("(a.nickname LIKE ? ESCAPE '\\' OR a.public_account_id LIKE ? ESCAPE '\\')");
const pattern = like(input.query.trim());
bindings.push(pattern, pattern);
}
if (input.ipLocation?.trim()) {
conditions.push("a.ip_location LIKE ? ESCAPE '\\'");
bindings.push(like(input.ipLocation.trim()));
}
if (input.cooperationSource?.trim()) {
conditions.push(
`EXISTS (SELECT 1 FROM distributions dx JOIN partners px ON px.id = dx.partner_id
WHERE dx.account_id = a.id AND px.name LIKE ? ESCAPE '\\')`,
);
bindings.push(like(input.cooperationSource.trim()));
}
if (input.platform?.trim() && input.platform !== "all") {
conditions.push("a.platform = ?");
bindings.push(input.platform.trim());
}
return { where: conditions.length ? `WHERE ${conditions.join(" AND ")}` : "", bindings };
}
export async function resourceSearch(input: ResourceFilters) {
await ensureSchema();
const { limit, offset } = page(input);
const { where, bindings } = resourceWhere(input);
const db = getRawDb();
const select = `SELECT a.*,
(SELECT COUNT(*) FROM distributions d WHERE d.account_id = a.id) AS cooperation_count,
(SELECT GROUP_CONCAT(DISTINCT p.name) FROM distributions d JOIN partners p ON p.id = d.partner_id WHERE d.account_id = a.id) AS cooperation_sources`;
const [rows, count] = await Promise.all([
db.prepare(`${select} FROM accounts a ${where} ORDER BY a.last_seen_at DESC LIMIT ? OFFSET ?`)
.bind(...bindings, limit, offset).all<Record<string, unknown>>(),
db.prepare(`SELECT COUNT(*) AS total FROM accounts a ${where}`).bind(...bindings).first<{ total: number }>(),
]);
return {
total: count?.total ?? 0,
limit,
offset,
accounts: rows.results.map((row) => ({
...row,
cooperation_sources: String(row.cooperation_sources ?? "").split(",").filter(Boolean),
})),
};
}
export async function resourceGet(accountId: string) {
await ensureSchema();
const db = getRawDb();
const account = await db.prepare("SELECT * FROM accounts WHERE id = ?").bind(accountId).first<Record<string, unknown>>();
if (!account) throw new Error("账号不存在");
const history = await db.prepare(
`SELECT d.id AS distribution_id, d.task_id, t.name AS task_name, c.title,
d.publish_url, d.publish_time, d.latest_likes, d.latest_comments, d.latest_collects,
d.exposure, d.views, p.name AS cooperation_source, cl.claimant_name
FROM distributions d JOIN tasks t ON t.id = d.task_id
JOIN contents c ON c.id = d.content_id JOIN partners p ON p.id = d.partner_id
LEFT JOIN claims cl ON cl.id = d.claim_id
WHERE d.account_id = ? ORDER BY COALESCE(d.publish_time, d.claimed_at) DESC`,
).bind(accountId).all<Record<string, unknown>>();
return { account, cooperation_history: history.results };
}
export async function backfillResourceProfile(
input: { distributionId?: string; publishUrl?: string },
bindings: McpOperationBindings,
) {
await ensureSchema();
const db = getRawDb();
const publishUrl = input.publishUrl ? extractXhsPublishUrl(input.publishUrl) : "";
const row = input.distributionId
? await db.prepare(
`SELECT d.id, d.publish_url, COALESCE(a.nickname, '') AS nickname
FROM distributions d LEFT JOIN accounts a ON a.id = d.account_id WHERE d.id = ?`,
).bind(input.distributionId).first<{ id: string; publish_url: string | null; nickname: string }>()
: publishUrl
? await db.prepare(
`SELECT d.id, d.publish_url, COALESCE(a.nickname, '') AS nickname
FROM distributions d LEFT JOIN accounts a ON a.id = d.account_id
WHERE d.publish_url = ? ORDER BY d.updated_at DESC LIMIT 1`,
).bind(publishUrl).first<{ id: string; publish_url: string | null; nickname: string }>()
: null;
if (!row?.publish_url) throw new Error("没有找到可补全的发布记录");
const result = await enrichDistributionAccount(
db,
row.id,
row.publish_url,
row.nickname,
resolveCollectionMcpConfig(bindings),
);
const account = result.updated && result.accountId
? await db.prepare("SELECT * FROM accounts WHERE id = ?").bind(result.accountId).first<Record<string, unknown>>()
: null;
return { distribution_id: row.id, ...result, account };
}
export async function resourceExport(input: ResourceFilters, origin: string) {
const result = await resourceSearch({ ...input, limit: 200, offset: 0 });
const allIds: string[] = result.accounts.map((account) => String(account.id));
if (result.total > result.accounts.length) {
const { where, bindings } = resourceWhere(input);
const rows = await getRawDb().prepare(`SELECT a.id FROM accounts a ${where} ORDER BY a.last_seen_at DESC LIMIT 5000`)
.bind(...bindings).all<{ id: string }>();
allIds.splice(0, allIds.length, ...rows.results.map((row) => row.id));
}
if (!allIds.length) throw new Error("当前筛选结果为空");
const issued = await issueMcpExportToken("resources", { accountIds: allIds });
return {
account_count: allIds.length,
download_url: `${origin}/api/resources-export?token=${encodeURIComponent(issued.token)}`,
expires_at: issued.expiresAt,
};
}

247
lib/mcp-tools.ts Normal file
View File

@@ -0,0 +1,247 @@
import { type McpServer } from "@modelcontextprotocol/server";
import { z } from "zod/v4";
import {
backfillResourceProfile,
collectNow,
recoveryExport,
recoveryList,
resourceExport,
resourceGet,
resourceSearch,
retryFailed,
runDueCollections,
setCollectionPlan,
taskGet,
taskList,
type McpOperationBindings,
} from "./mcp-operations";
type Options = {
bindings: McpOperationBindings;
origin: string;
};
function result(label: string, data: unknown) {
return {
content: [{ type: "text" as const, text: label }],
structuredContent: data as Record<string, unknown>,
};
}
function errorResult(error: unknown) {
const message = error instanceof Error ? error.message : "操作失败,请稍后重试";
return {
isError: true as const,
content: [{ type: "text" as const, text: message.slice(0, 240) }],
};
}
function withError<T extends unknown[]>(handler: (...args: T) => Promise<ReturnType<typeof result>>) {
return async (...args: T) => {
try {
return await handler(...args);
} catch (error) {
return errorResult(error);
}
};
}
const pagination = {
limit: z.number().int().min(1).max(200).optional().describe("返回数量默认50最大200"),
offset: z.number().int().min(0).optional().describe("分页偏移量默认0"),
};
const resourceFilters = {
query: z.string().max(100).optional().describe("账号名称或小红书号/抖音号,支持模糊搜索"),
ip_location: z.string().max(100).optional().describe("IP地区关键词支持模糊搜索"),
cooperation_source: z.string().max(100).optional().describe("历史合作来源关键词,支持模糊搜索"),
platform: z.string().max(30).optional().describe("平台,例如小红书;不传表示全部"),
};
export function registerMcpOperationTools(server: McpServer, options: Options) {
server.registerTool(
"task_list",
{
title: "查询任务及进度",
description: "查询 KOC 分发任务列表、领取数、发布数、第7天回收数及领取链接。",
inputSchema: z.object({
query: z.string().max(100).optional().describe("任务名或品牌/项目关键词"),
status: z.string().max(30).optional().describe("任务状态;不传或 all 表示全部"),
...pagination,
}),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: true },
},
withError(async (input) => {
const data = await taskList(input, options.bindings.KOC_PORTAL_URL?.trim() || "");
return result(`共找到 ${data.total} 个任务。`, data);
}),
);
server.registerTool(
"task_get",
{
title: "查看任务完整情况",
description: "查看指定任务、笔记、领取记录、发布回填和采集执行记录。",
inputSchema: z.object({ task_id: z.string().min(1).describe("任务ID") }),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: true },
},
withError(async ({ task_id }) => {
const data = await taskGet(task_id, options.bindings.KOC_PORTAL_URL?.trim() || "");
return result(`已读取任务“${String(data.task.name)}”的完整情况。`, data);
}),
);
server.registerTool(
"recovery_list",
{
title: "查询数据回收队列",
description: "查询已发布、未回填、待第7天数据或已到第7天仍未回填的笔记。",
inputSchema: z.object({
task_id: z.string().optional().describe("任务ID不传则跨任务查询"),
stage: z.enum(["all", "published", "unfilled", "waiting_day7", "day7_due"]).optional(),
...pagination,
}),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: true },
},
withError(async ({ task_id, stage, limit, offset }) => {
const data = await recoveryList({ taskId: task_id, stage, limit, offset });
return result(`数据回收队列共 ${data.total} 条记录。`, data);
}),
);
server.registerTool(
"recovery_export",
{
title: "导出任务完整数据",
description: "生成任务完整 Excel包含笔记原图、发布截图、创作者截图和回收数据。",
inputSchema: z.object({ task_id: z.string().min(1).describe("任务ID") }),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: false },
},
withError(async ({ task_id }) => {
const data = await recoveryExport(task_id, options.origin);
return result(`导出文件已生成,下载链接将在 ${data.expires_at} 失效。`, data);
}),
);
server.registerTool(
"collection_plan_set",
{
title: "设置自动采集计划",
description: "为任务设置开始日期及第1至第7天的自动采集日系统在北京时间10:00执行。",
inputSchema: z.object({
task_id: z.string().min(1),
start_date: z.string().regex(/^\d{4}-\d{2}-\d{2}$/).describe("北京时间开始日期 YYYY-MM-DD"),
days: z.array(z.number().int().min(1).max(7)).min(1).max(7).describe("需要采集的相对天数,例如 [2,5,7]"),
}),
annotations: { readOnlyHint: false, destructiveHint: false, idempotentHint: true, openWorldHint: true },
},
withError(async ({ task_id, start_date, days }) => {
const data = await setCollectionPlan(task_id, start_date, days, options.bindings);
return result(`已为任务设置 ${data.days.length} 个自动采集日。`, data);
}),
);
server.registerTool(
"collection_run_due",
{
title: "执行到期采集",
description: "立即执行今天或此前已到期但尚未成功的自动采集任务。",
inputSchema: z.object({ task_id: z.string().optional().describe("任务ID不传则执行所有到期任务") }),
annotations: { readOnlyHint: false, destructiveHint: false, idempotentHint: true, openWorldHint: true },
},
withError(async ({ task_id }) => {
const data = await runDueCollections(task_id, options.bindings);
return result(`到期采集完成:成功 ${data.succeeded},失败 ${data.failed}`, data);
}),
);
server.registerTool(
"collection_collect_now",
{
title: "立即采集指定笔记",
description: "对指定分发记录立即采集点赞、收藏和评论数据。",
inputSchema: z.object({
distribution_id: z.string().min(1).describe("分发记录ID可从 task_get 或 recovery_list 获取"),
schedule_day: z.number().int().min(1).max(7).optional().describe("标记为第几天采集,可不传"),
}),
annotations: { readOnlyHint: false, destructiveHint: false, idempotentHint: true, openWorldHint: true },
},
withError(async ({ distribution_id, schedule_day }) => {
const data = await collectNow(distribution_id, schedule_day, options.bindings);
return result("指定笔记采集完成。", data);
}),
);
server.registerTool(
"collection_retry_failed",
{
title: "补采异常数据",
description: "重试指定任务或全部任务中采集状态异常的已发布笔记。",
inputSchema: z.object({ task_id: z.string().optional().describe("任务ID不传则补采最近异常任务") }),
annotations: { readOnlyHint: false, destructiveHint: false, idempotentHint: true, openWorldHint: true },
},
withError(async ({ task_id }) => {
const data = await retryFailed(task_id, options.bindings);
return result(`补采完成:成功 ${data.succeeded},失败 ${data.failed}`, data);
}),
);
server.registerTool(
"resource_search",
{
title: "搜索 KOC 账号资源",
description: "按账号名称/账号号、IP地区、合作来源或平台搜索 KOC 资源。",
inputSchema: z.object({ ...resourceFilters, ...pagination }),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: true },
},
withError(async ({ query, ip_location, cooperation_source, platform, limit, offset }) => {
const data = await resourceSearch({ query, ipLocation: ip_location, cooperationSource: cooperation_source, platform, limit, offset });
return result(`共找到 ${data.total} 个账号。`, data);
}),
);
server.registerTool(
"resource_get",
{
title: "查看 KOC 账号详情",
description: "查看账号主页、账号号、粉丝数、IP地区以及全部合作记录。",
inputSchema: z.object({ account_id: z.string().min(1).describe("KOC LOOP 账号ID") }),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: true },
},
withError(async ({ account_id }) => {
const data = await resourceGet(account_id);
return result(`已读取账号“${String(data.account.nickname)}”的详情。`, data);
}),
);
server.registerTool(
"resource_backfill_profile",
{
title: "补全公开账号信息",
description: "根据已回填的发布链接补全账号主页、昵称、小红书号、IP地区和粉丝数。",
inputSchema: z.object({
distribution_id: z.string().optional().describe("分发记录ID和发布链接二选一"),
publish_url: z.string().optional().describe("小红书发布链接或包含链接的分享文案"),
}).refine((value) => Boolean(value.distribution_id || value.publish_url), "请提供分发记录ID或发布链接"),
annotations: { readOnlyHint: false, destructiveHint: false, idempotentHint: true, openWorldHint: true },
},
withError(async ({ distribution_id, publish_url }) => {
const data = await backfillResourceProfile({ distributionId: distribution_id, publishUrl: publish_url }, options.bindings);
return result(data.updated ? "账号公开信息已补全。" : "账号信息未发生变化。", data);
}),
);
server.registerTool(
"resource_export",
{
title: "导出 KOC 资源",
description: "按账号、IP地区、合作来源或平台筛选并导出 KOC 资源 Excel。",
inputSchema: z.object(resourceFilters),
annotations: { readOnlyHint: true, destructiveHint: false, idempotentHint: false },
},
withError(async ({ query, ip_location, cooperation_source, platform }) => {
const data = await resourceExport({ query, ipLocation: ip_location, cooperationSource: cooperation_source, platform }, options.origin);
return result(`已生成 ${data.account_count} 个账号的导出文件。`, data);
}),
);
}

View File

@@ -165,6 +165,13 @@ export async function ensureSchema(database?: D1Database) {
expires_at TEXT NOT NULL,
created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP
)`,
`CREATE TABLE IF NOT EXISTS mcp_export_tokens (
token_hash TEXT PRIMARY KEY,
kind TEXT NOT NULL,
payload TEXT NOT NULL,
expires_at TEXT NOT NULL,
created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP
)`,
];
for (const statement of statements) {
@@ -340,6 +347,11 @@ export async function ensureSchema(database?: D1Database) {
"CREATE INDEX IF NOT EXISTS auth_sessions_expires_at_idx ON auth_sessions(expires_at)",
)
.run();
await db
.prepare(
"CREATE INDEX IF NOT EXISTS mcp_export_tokens_expires_at_idx ON mcp_export_tokens(expires_at)",
)
.run();
await db
.prepare(
`UPDATE distributions